|
Activity Number:
|
202
- Meta-Analysis, Mediation, and Causal Inference from a Bayesian Perspective
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 8, 2022 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
|
Abstract #322320
|
|
|
Title:
|
A Bayesian Comparative Effectiveness Trial in Action: Executing a Multi-Site Study with Response-Adaptive Randomization
|
|
Author(s):
|
Alexandra Brown* and Byron Gajewski and Dinesh Pal Mudaranthakam and Mamatha Pasnoor and Mazen Dimachkie and Omar Jawdat and Laura Herbelin and Richard Barohn
|
|
Companies:
|
University of Kansas Medical Center and University of Kansas Medical Center and University of Kansas Medical Center and University of Kansas Medical Center and University of Kansas Medical Center and University of Kansas Medical Center and University of Kansas Medical Center and University of Missouri Health Care
|
|
Keywords:
|
Bayesian design;
adaptive;
response adaptive randomization;
comparative effectiveness;
interim analyses;
posterior probabilities
|
|
Abstract:
|
PAIN-CONTROLs was a comparative effectiveness study utilizing a Bayesian adaptive design with response adaptive randomization to one of four drugs for cryptogenic sensory polyneuropathy (CPSN) patients. The outcome of the study was a utility function that combined efficacy and quit rates for each drug at endpoint. We calculated the posterior probability that a drug was best (pp) by comparing utility functions of all four drugs. After a ‘burn-in’, an interim analysis that provided new allocation probabilities was performed quarterly. The updated probabilities were driven by the drug’s sample size and performance. The trial could stop early after 100 patients had endpoint data and if the pp > 0.925 (success). Six interim analyses were performed until the maximum accrual was reached. The pp fluctuated for nortriptyline and duloxetine being the best with fewer patients being allocated to pregabalin and mexiletine. At the final analysis, a drug was best if pp > 0.925 or a “loser” if pp < 0.01. While there was no best, nortriptyline and duloxetine outperformed pregabalin and mexiletine. Mexiletine is defined as a loser, the probability it is the best was < 0.01.
|
Authors who are presenting talks have a * after their name.
